Recent models have introduced the notion of dimensions and hierarchies in social networks. These models motivate the mining of small world graphs under a new perspective. We exemplary base our work on a conference graph, which is constructed from the DBLP publication records. We show that this graph indeed exhibits a layered structure as the models suggest. We then introduce a subtraction approach that allows to segregate layers. Using this technique we separate the conference graph into a thematic and a quality layer. As concrete applications of the discussed methods we present a novel rating method as well as a conference search tool that bases on our graph and its layer separation.